AI Foundations Research Intern: Build Multimodal AI

IBM; Cambridge (MA); On-site; USD 41,000 - 55,000; Part time

Job summary

IBM Research in Cambridge, MA seeks an AI Research Scientist Intern to contribute to foundations of AI, including multi-modal models, training algorithms, and governance. This role offers exposure to cutting-edge research and collaboration with a diverse team.

You will design experiments, prototype solutions in Python, and publish results, blending fundamental science with engineering. A Master’s degree is required, and a strong publication record is preferred.

Qualifications

• Hands-on experience with LLMs, VLMs and Transformer architectures.
• Proficiency in Python for rapid prototyping and ML systems.
• Strong analytical and problem-solving abilities.
• Proven research and communication skills with publications or preprints.
• Collaborative, reproducible research practices.
• Experience with ML frameworks (PyTorch) and end-to-end algorithm development.
• Experience with generative AI and multimodal models.

Responsibilities

• Engage in the full research lifecycle from ideation to publication.
• Design novel prototype solutions and validate them through experiments.
• Conduct foundational research in training algorithms and RL.
• Collaborate in mentored teams to shepherd projects to completion.
• Disseminate results through publications and patent applications.

Skills

LLMs

VLMs

Transformer architectures

Python prototyping

Experimental design

Research communications

Collaborative research

Education

Master's Degree

Tools

PyTorch

Python

Job description

IBM Research in Cambridge, MA seeks an AI Research Scientist Intern to contribute to foundations of AI, including multi-modal models, training algorithms, and governance. This role offers exposure to cutting-edge research and collaboration with a diverse team.

You will design experiments, prototype solutions in Python, and publish results, blending fundamental science with engineering. A Master’s degree is required, and a strong publication record is preferred.